-
XIAOYAN DENG, BARBARA E. KAHN, H. RAO UNNAVA, and HYOJIN
LEE*
The authors investigate how horizontal versus vertical displays
ofalternatives affect assortment processing, perceived variety, and
subsequentchoice. Horizontal (vs. vertical) displays are easier to
process due to a matchbetween the human binocular vision field
(which is horizontal in direction) andthe dominant direction of eye
movements required for processing horizontaldisplays. It is
demonstrated that this processing fluency allows people tobrowse
information more efficiently, which increases perceived
assortmentvariety and ultimately leads to more variety being
chosen, and if the number ofoptions chosen is allowed to vary, it
leads to more options chosen. It is shownthat because people see
more variety in a horizontal (vs. vertical) display,they process a
horizontal assortment more extensively. When more varietyis
positive, they find the choice task easier and have a higher level
ofsatisfaction and confidence about their choices. When more
variety is notnecessarily positive, for example, in a choice of a
single most-preferredoption, these effects disappear. Two field
studies, an eye-tracking study,and two lab studies support these
conclusions.
Keywords: horizontal versus vertical, retail assortment,
perceived variety,eye movement, variety seeking
Online Supplement : http:dx.doi.org/10.1509/jmr.13.0151
A “Wide” Variety: Effects of Horizontal VersusVertical Display
on Assortment Processing,Perceived Variety, and Choice
Assortment variety is one of the critical variables thatattract
consumers to a store or to an online retail website(Iyengar and
Lepper 2000; Mantrala et al. 2009). However,research has shown that
despite consumers’ preference formore variety when choosing among
assortments, larger se-lections can yield lower purchase
probabilities because ofchoice overload (Iyengar and Lepper 2000).
One approach tominimizing the effect of too much variety is to
manage the
perceived rather than the actual variety (Broniarczyk, Hoyer,and
McAlister 1998). Broniarczyk, Hoyer, and McAlister(1998) find that
they can reduce SKUs without loweringperceived variety. Consistent
with this finding, Townsendand Kahn (2014) show that even if the
actual variety is heldconstant, higher perceptions of variety can
attract consumers,making them more likely to choose; Kahn and
Wansink(2004) find that higher levels of perceived variety can
causepeople to consume more.
Perceived variety can be increased by managing struc-tural
aspects of the assortment. For example, Morales et al.(2005) show
that if the retailer designs the external structureof a product
category to match consumers’ internal catego-rization of that
product category, consumers will perceivemore variety and will be
more satisfied with their choices.Mogilner, Rudnick, and Iyengar
(2008) show that the merepresence of category labels within a
product assortment canincrease satisfaction with that assortment,
especially for un-familiar products. Townsend and Kahn (2014) show
that for
*Xiaoyan Deng is Assistant Professor of Marketing, Fisher
College ofBusiness, The Ohio State University (e-mail:
[email protected]). Barbara E.Kahn is Patty and Jay H. Baker
Professor; Director, Jay H. Baker RetailingCenter; and Professor of
Marketing, The Wharton School, University ofPennsylvania (e-mail:
[email protected]). H. Rao Unnava is Deanand Professor,
Graduate School of Management, University of California,Davis
(e-mail: [email protected]). Hyojin Lee is Assistant Professor
ofMarketing, Lucas College and Graduate School of Business, San
Jose StateUniversity (e-mail: [email protected]). Joel Huber
served as guest editorand Michel Wedel served as associate editor
for this article.
© 2016, American Marketing Association Journal of Marketing
ResearchISSN: 0022-2437 (print) Ahead of Print
1547-7193 (electronic) DOI: 10.1509/jmr.13.01511
http:dx.doi.org/10.1509/jmr.13.0151mailto:[email protected]:[email protected]:[email protected]:[email protected]://dx.doi.org/10.1509/jmr.13.0151
-
large assortments, design features (e.g., visual vs.
verbaldepiction) can be used to motivate consumers to spendmore
time processing and to process the options withinthe assortment
more systematically, which leads to lessperceived complexity and
mitigates delay in choosing.
One of the critical ways consumers form assortmentvariety
perceptions is by visually scanning the offering(Broniarczyk,
Hoyer, and McAlister 1998; Hoch, Bradlow,and Wansink 1999); thus,
it is very important that retailersgrab the consumers at this
visceral visual level. A funda-mental visual cue that is widely
used by merchants to attractattention, direct movement, and
influence behavior is theuse of horizontal versus vertical displays
(see Figure 1)or the use of horizontal versus vertical blocking
(seeFigure 2). Blocking divides the display into horizontal
orvertical blocks of similar products; the grouping can bebased on
product subcategory (Figure 2, Panel A), brand(Figure 2, Panel B),
or product attributes (Figure 2, Panels Cand D). It is almost
always achieved through packagingdesign, especially by use of a
distinct packaging color foreach grouping. Retailers use color cues
to visually directcustomers’ eye movements, thereby helping them
navi-gate through a cluttered store shelf (DuPuis and
Silva2008).
We hypothesize that this visual factor (horizontal vs. ver-tical
display or blocking) will influence how consumersprocess an
assortment, which in turn will affect howmuch variety they perceive
and subsequently how muchvariety they choose. Horizontal displays
should be easierto process than vertical displays due to a match
betweenbinocular vision field (which is horizontal) and the
domi-nant direction of eye movements for processing hori-zontal
displays. This ease of processing should allowpeople to take in
more information about the assortment
in a given time, which should increase perceptions ofvariety.
Higher levels of perceived variety should leadto more variety
chosen (Kahn and Wansink 2004). Fur-thermore, because people see
more variety in a horizontalversus a vertical display, when time is
unconstrained, theyshould process the assortment more extensively.
Whenthe choice task allows for multiple options or multiplevariants
to be chosen, these characteristics of horizontal ori-entation
should make the choice task easier (Townsendand Kahn 2014) and
should lead to higher levels of sat-isfaction and confidence in
their choices (Hoch, Bradlow,and Wansink 1999). In choice tasks
wherein more varietyis not necessarily positive, such as a task to
choose onlyone favorite or choices for which simplicity and
clearstructure is preferred, the advantages of horizontal
ori-entation disappear.
We begin with two field experiments (Studies 1–2) thatprovide
support for our proposition that all else being heldequal,
horizontal (vs. vertical) displays will lead to morevariety being
chosen, that is, more unique items chosen.Then, in a lab experiment
(Study 3), we show that when wetightly constrain the amount of time
for processing an as-sortment, horizontal (vs. vertical) displays
result in higherperceptual fluency. This initial processing fluency
leadsto perceptions of higher variety for the horizontal
displayseven when longer time is allowed for scanning,
presumablybecause once perceptions are formed, people often use
themto make subsequent judgments, do not seek
disconfirmingevidence, and even interpret later information around
theseinitial impressions (Tversky and Kahneman 1974). These
in-creases in perceived variety in turn result in more unique
op-tions chosen for the horizontal displays.
In a subsequent eye-tracking study (Study 4), we findevidence
that in constrained-time conditions, participants
Figure 1HORIZONTAL VERSUS VERTICAL RETAIL ASSORTMENT
DISPLAYS
Horizontal
Vertical
2 JOURNAL OF MARKETING RESEARCH, Ahead of Print
-
are able to process more items per second of processingtime in
the horizontal versus vertical conditions. They alsoform larger
consideration sets in the horizontal conditions.
Later, when they see the same assortments again with un-limited
time, they spend more time and process moreitems in the horizontal
versus vertical displays, presumably
Figure 2HORIZONTAL VERSUS VERTICAL BLOCKING
Hea
lth
y &
Del
icio
us
Cla
ssic
Fav
ori
tes
Tast
e S
ensa
tio
ns
Fu
n F
avo
rite
s
A: Sorted by Subcategory C: Sorted by Product Attributes
B: Sorted by Brand D: Sorted by Product Attributes
Notes: Panels A and B are examples of vertical blocking, and
Panels C and D are examples of horizontal blocking.
Effects of Horizontal Versus Vertical Display 3
-
because they started with higher initial perceptions of
as-sortment variety and larger consideration sets for the
hori-zontal displays. This appears to lead to larger choice sets
forthe horizontal displays.
Finally, in a quasi field study (Study 5), we test a
boundarycondition for our effects. In all of our previous
experi-ments, participants were allowed to choose multiple op-tions
from the assortment. In this study, using familiarstimuli (branded
chocolate candies), we add a conditionin which participants may
only chose one option. In thissingle-choice condition, perceived
variety is less relevantbecause consumers are focusing on choosing
their favoriteoption. We show that in the multiple-choice
condition, con-sumers spend more time processing the assortment,
findthe choice task easier, and have greater satisfaction
andconfidence about their choices for horizontal versus ver-tical
displays. These effects are not evident in the single-choice
condition.
Our findings have important managerial implicationsbecause
marketers seeking a customer share advantagehave turned
increasingly to managing the visual impact ofpoint-of-purchase
displays. Assortment variety is amongthe top three drivers of
retail patronage, along with lo-cation and price (Broniarczyk and
Hoyer 2010). Ourfindings suggest that when variety is positive
(e.g., notcontributing to choice overload), assortments designed
forhorizontal (vs. vertical) scanning can promote
increasedperceptions of assortment variety. Higher perceived
assort-ment variety can increase share of wallet and consump-tion
by increasing purchase quantities (Kahn and Wansink2004).
We begin with the results of two field studies that doc-ument
the central proposition that horizontal (vs. vertical)displays
increase the variety of options chosen.
STUDY 1: SHOPPING MALL FIELD STUDY
We conducted a field study on a Saturday in February ina large
U.S. city, in a national chain retail store that spe-cializes in
bath items, personal care items, and home fra-grances (see Figure
3). The store used two types of fixture toshowcase their
pocket-sized hand sanitizers: a horizontaltable display and a
vertical display named “bubble tree”(see Figure 3). These two
displays served as our manipulation.In this study, sample size was
determined by recruiting asmany participants as possible in one day
in the store, witha minimum requirement of 20 participants per
condition(Simmons, Nelson, and Simonsohn 2011). No compensationwas
offered to the participants.
For our study, we placed the two types of display fixtureat two
different, but equally visible and accessible, loca-tions in the
store. Each fixture contained 32 bins to display32 different
fragrances of the hand sanitizer. Each bin wasfilled with an equal
number of items. Although the photographsin Figure 3 indicate that
there were different promotional signsin the different displays,
all signs were actually removed inthe experimental tests.
When customers purchased this product, a store associateasked
them to fill out a short survey about their shoppingexperience. In
the survey, customers first checked the fra-grances they chose (on
a list of the 32 displayed fragrances)and reported the purchase
quantity for each checked fra-grance. Then, they indicated whether
they shopped from the
horizontal or from the vertical display (those who browsedor
shopped both displays were not asked to do the survey).Finally,
they provided an overall evaluation of the specificstore display (1
= “bad,” and 9 = “good”), as well as somepersonal information.
Sixty-seven customers (age 11–63; mean age = 36; 64females)
completed the survey. One-way (horizontal vs.vertical display)
between-subjects analyses of variance(ANOVAs) showed that, compared
with those who shoppedthe vertical display, customers who shopped
the horizon-tal display purchased a greater quantity (Mhorizontal =
4.44,standard error of the mean [SEM] = .44 vs. Mvertical =
2.75,SEM = .58; F(1, 65) = 5.4, p = .02, h2 = .08) as well asa
greater variety (number of distinct fragrances in thechoices:
Mhorizontal = 3.67, SEM = .29 vs. Mvertical = 2.38,SEM = .38; F(1,
65) = 7.4, p = .008, h2 = .10). There wasno difference in the
evaluation of the two store displays(Mhorizontal = 7.88, SEM = .14
vs. Mvertical = 8.00, SEM =.19; F(1, 65) = .2, p = .63, h2 = .004).
We also found thatthe number of different fragrances purchased
(i.e., amountof variety seeking) mediated the total quantity
purchased.Our bootstrap analysis indicated that the indirect
effectof display orientation (horizontal vs. vertical) on
purchasequantity through the hypothesized mediator, variety
seeking,was significant (b = .27, SE = .11; 95% confidence
interval[CI] = [.08, .51]).
This field experiment provides support with actual pur-chase
behavior that horizontal store displays, as opposedto vertical
displays, promote a more diverse set of chosenoptions, resulting in
increased purchase quantities. Next,we report the results of a
second field study on a differ-ent population (children), wherein
the dependent variable ischoice rather than purchase.
STUDY 2: HALLOWEEN FIELD STUDY
The Halloween tradition of having children come to housesand
choose candy from an assortment offered us the oppor-tunity to
develop a simple experiment to test how horizontalversus vertical
displays affect the amount of variety chosen. Inthis study, sample
size was determined by recording data foras many participants as
possible during the “trick or treat”hours, with a minimum
requirement of 20 participants percondition. No compensation was
offered to the participants.Each child was offered an assortment of
the following fivecandies: Hershey’s, KitKat, M&M’s, Snickers,
and Twix.Each of the five candies was presented in bulk to the
childrenin five transparent containers that could be arranged on
atable either vertically or horizontally (see Figure 4).
Becauseeach container’s height is equal to its width, the
horizontaland vertical arrays of candy bins were of equal length.
Wepresented the candy types in two random orders, which
werecounterbalanced across participants. When
trick-or-treaters(approximate ages ranging from 3 to 14 years old)
came tothe door, they were assigned to either the vertical or the
hor-izontal condition. In each case, the children were told that
theycould have any three pieces of candy. For example, they
couldhave three different candies, three of the same, or any
com-bination that added up to three. The candy containers
wererefilled so that every time a child approached the
containers,they all appeared equally full. The study was
videotaped,and we later replayed the video to record and analyze
whichcandies each child picked.
4 JOURNAL OF MARKETING RESEARCH, Ahead of Print
-
Forty-one children came to the house between 6:00 and7:30 P.M.
Of these children, we excluded seven from the data(three children
picked fewer than three candies, three childrenpicked more than
three, and one child started pickingcandies before receiving the
instructions), resulting in afinal sample size of 34. The
horizontal condition had 16children and the vertical condition had
18. Due to the smallsample size, several cells of the six
possibilities (2 [orders:horizontal vs. vertical] × 3 [number of
unique items chosen:one, two, or three]) had fewer than five
participants per cell,so we focused only on whether children chose
the maximumlevel of variety or did not. In the horizontal
condition, 13 ofthe 16 children (81%) chose three unique candies.
In the verticalcondition, 8 of the 18 children (44%) chose three
uniquecandies. Of the 21 children who chose three unique
candies,
13 (62%) were in the horizontal condition, and 8 (38%) werein
the vertical condition. Either way it is sorted, a
significantdifference exists between the vertical and horizontal
con-ditions (c2(1, 34) = 4.9, p = .03).
Both of these field studies provide evidence that hori-zontal
displays result in the choice of more variety thanvertical
displays. We propose that horizontal displays leadto the choice of
more variety because they can be processedmore efficiently than
vertical displays, resulting in greaterperceived variety. The
perceptions of higher variety appearto lead to larger consideration
sets, which eventually lead tolarger and more varied choice
sets.
We now discuss how information displayed horizontallyversus
vertically is processed by the human eye to result inthe
differences detected in the first two field studies. To test
Figure 3STUDY 1 (SHOPPING MALL FIELD STUDY) MANIPULATION AND
STIMULI
Horizontal Display Vertical Display
Notes: All promotional signs were removed during the study. The
number of bins (32) was held constant across the horizontal and
vertical displays, and each binwas filled with an equal number of
items.
Effects of Horizontal Versus Vertical Display 5
-
our theorization, we then turn to the laboratory for our
nextthree experiments (see Figure 5 for a road map of these
threestudies).
PROCESSING OF INFORMATION IN HORIZONTALVERSUS VERTICAL
DISPLAYS
When consumers are faced with a product display, theyvisually
process the assortment in amatter of seconds (Dicksonand Sawyer
1990; Hoyer 1984). We propose that this is easierif the assortment
is scanned horizontally versus vertically,primarily because
horizontal eye movements result in higherprocessing fluency of
assortment information because ourfield of view is wider in the
horizontal versus vertical di-rection. Humans view the world
through binocular visionbecause they use two eyes in unison. As
Figure 6, Panel A,shows, humans’ forward-facing eyes give a maximum
hori-zontal field of view of about 190° (120° make up the field
ofview seen by both eyes, shown in gray in Figure 6, Panel
A,whereas the white area on each side of the figure indicates
theleft or right eye’s unique field of vision) (Academic
Medicine1960). The design of televisions, computermonitors, and
otherscreen-based products closely mimics the shape of
binocularvision presumably because such a shape facilitates the
pro-cessing of information displayed on the screen (Metallinos1996;
see Figure 6, Panel B).
In addition, the perceptual span is the area from which
anindividual acquires information during a fixation (Rayner1975).
Reading studies have shown that the horizontal spanof this area is
almost twice as large as the vertical span(Ojanpää, Näsänen,
and Kojo 2002). The perceptual span invisual processing mirrors
that for reading (Rayner 1998).That the perceptual span is wider in
the horizontal direc-tion should facilitate horizontal (vs.
vertical) eye movement,because there is more information available
to direct the
eyes horizontally (Shi,Wedel, and Pieters 2013). This
suggeststhat horizontal scanning should be more fluent than
verticalscanning.
Horizontal scanning is also easier because the
musclescontrolling horizontal eye movements are stronger than
thosecontrolling verticalmovement (Cogan 1956). Finally, our
headnaturally rests forward (Griegel-Morris et. al 1992), so it
takesmore muscular effort to lift our eyes and scan vertically.This
suggests that making vertical eye movements is moreeffortful and
less convenient.
Based on the preceding points, we first conclude that
hori-zontal scanning of information is more fluent than
verticalscanning. The observation of the dominance of horizontal
eyemovements in various visual tasks (Gilchrist and Harvey2006;
Tatler and Vincent 2008; Van der Lans, Pieters andWedel 2008)
supports this conclusion. Also supporting ourconclusion, various
studies have found that performancein visual attention tasks is
significantly better on thehorizontal dimension than on the
vertical dimension ofthe visual field (Collewijn, Erkelens, and
Steinman 1988;Kröse and Julesz 1989; Nazir 1992; Yeshurun and
Carrasco1999).
Second, because horizontal scanning is more fluent thanvertical
scanning, we predict a higher level of processing ef-ficiency when
assortments are designed to promote horizontalversus vertical
scanning. This is consistent with the finding thatthe subjective
experience of perceptual fluency is based on theobjective
processing speed (Reber, Wurtz, and Zimmermann2004). The common
practice used to assess perceptual fluencyand processing speed is
to control the stimulus exposure time(Reber, Schwarz, and
Winkielman 2004). We follow thispractice to test the processing
advantages of horizontal (vs.vertical) assortments in two lab
experiments in which weconstrain processing time (between subjects
in Study 3 and
Figure 4STUDY 2 (HALLOWEEN FIELD STUDY) MANIPULATION AND
STIMULI
Horizontal Display Vertical Display
6 JOURNAL OF MARKETING RESEARCH, Ahead of Print
-
within subject in Study 4). In Study 3, we measure thesubjective
perceived fluency of processing and find supportfor horizontal
advantages in the constrained-time condition.In Study 4, using
eye-tracking technology, we show thathorizontal processing is more
efficient, resulting in moreoptions fixated per second in
constrained-time scans of anassortment.
The processing fluency and efficiency associated withhorizontal
assortments should allow consumers to take in alarger amount of
information in a given, limited time. Thisshould lead them to
conclude that horizontal (vs. vertical)assortments have a higher
level of variety. Although theprocessing advantages in terms of
amount of informationprocessed might dissipate when people have
more time toprocess, we believe the initial perceptions of higher
varietycould still frame subsequent decisions. We hypothesize
thisfor several reasons: (1) once people form perceptions,
theyoften use them to make subsequent judgments; (2) they arenot
likely to seek disconfirming evidence although there
areopportunities to do so; and (3) they may even distort
laterinformation in a way that is consistent with the initial
per-ceptions (Kardes 1986; Russo et al. 2008; Tversky andKahneman
1974). Thus, these initial perceptions of greatervariety should
linger, resulting in the formation of largerconsideration sets in
horizontal (vs. vertical) displays.
We would then hypothesize that in choice situations inwhich time
is unconstrained, participants would spend alonger time processing
the horizontal (vs. vertical) assortmentand fixate on a higher
number of items because these horizontalassortments are associated
with perceptions of greater varietyand larger consideration sets.
However, because the horizontalprocessing advantages dissipate with
time, we would not ex-pect to see the higher processing
efficiencies (i.e., more optionsfixated on per second) linger when
more time is available—wehypothesize only that the initial
perceptions of variety linger.
These conclusions that higher perceived variety wouldyield
larger consideration sets support previous researchthat has shown
higher perceived assortment variety can
increase likelihood of purchase as well as consumption andchoice
quantities (Broniarczyk, Hoyer, andMcAlister 1998;Kahn and Wansink
2004). Thus, we predict that becauseconsumers perceive more variety
from horizontal (vs. vertical)displays of product assortments, they
are also likely to choosemore items.
To summarize, a horizontal display, because of its easeof
processing, leads people to process greater variety in agiven time,
and that perception of greater variety in turnleads them to choose
more variety. We test this causal linkin our next lab study, Study
3.
STUDY 3: LAB STUDY MEASURING PROCESSINGFLUENCY AND PERCEIVED
VARIETY
Method
Study 3 was computer mediated and used a 2
(assortmentorientation: horizontal vs. vertical) × 2 (exposure
time: 3sec vs. 15 sec) between-subjects design. Our sample was
215undergraduate students who participated in the study inexchange
for course credit. Each participant was randomlyassigned to one of
the four conditions. Sample size wasdetermined by recruiting as
many participants as possiblein one day in our lab, with a minimum
requirement of 20participants per condition.
We told participants to view an assortment of chocolatesas if
they were in a candy store. We showed them either ahorizontal or a
vertical display of 10 chocolate truffles (seeFigure 7) for either
3 or 15 sec. The exposure time ma-nipulation was based on the
assumption that if horizontalprocessing fluency advantages existed,
we would need avery tight time frame to observe them (Reber,
Schwarz, andWinkielman 2004; Reber, Wurtz, and Zimmermann
2004).Previous research that has shown consumers often spendonly a
brief time before making purchases decisions (Dicksonand Sawyer
1990; Hoyer 1984) lends support for the eco-logical validity of
this time manipulation. Although theseinitial processing advantages
likely dissipate over time, we
Figure 5OVERVIEW OF STUDY 3, STUDY 4 (EYE-TRACKING STUDY), AND
STUDY 5 (QUASI FIELD STUDY)
Horizontal (vs. vertical)assortment display
Processing fluency(dissipates with greater
processing time)
Perceived variety(lingers even when more
time is available)
Variety seeking(number of unique
items chosen)
Processing efficiency(number of items fixated
on per second ofprocessing)
Processing extensiveness(total processing time and
total number of items fixatedon before making choices)
Choice ease
Choice satisfaction
Choice confidence
Study 3
Study 4
Study 5
Effects of Horizontal Versus Vertical Display 7
-
are interested in whether the perceptions of higher
varietylinger, leading to differences in subsequent choice
decisions.
Regardless of orientation, both displays fit within thecomputer
screen, so participants did not have to scroll left–right or
up–down to view all the items. We presented the 10chocolates in the
display in two random orders, which werecounterbalanced across
participants.
After viewing the chocolate assortment, participants respon-ded
to a series of questions. First, perceived assortment varietywas
measured by a nine-point scale: “How much variety doyou think there
is in this assortment?” (1 = “very little variety,”and 9 = “very
much variety”). This measure is consistent withprevious research
(Broniarczyk, Hoyer, and McAlister 1998;Hoch, Bradlow, andWansink
1999; Kahn andWansink 2004).Second, number of unique options chosen
was measuredas follows: “Consider choosing seven chocolates from
this
assortment. You can choose either seven of the same flavor,or up
to seven different flavors. How many different flavorswould you
like to choose?” (1 = “1 flavor in 7 chocolates,”and 7 = “7 flavors
in 7 chocolates”). Third, perceived pro-cessing fluency was
measured using two items rated on nine-point scales: “How easy was
it to evaluate the informationshown in this assortment?” (1 = “not
at all easy,” and 9 =“very easy”) and “How comfortable did you feel
when youevaluated the information shown in this assortment?” (1
=“not at all comfortable,” and 9 = “very comfortable”). Theseitems
were adapted from Lee and Aaker (2004) and com-bined to form a
“processing fluency” index (a = .80).
Results and Discussion
We expected to observe a horizontal processing fluencyadvantage
only in the 3-sec condition, in which partici-pants had to process
the assortment under time constraint.In the 15-sec condition, in
which participants had more thanenough time to survey the
assortment options, we expected no
Figure 6BINOCULAR VISION FIELD
B: Binocular Vision Field Influences Television Designs
A: Binocular Vision Field Fitted with Internal and External
Rectangles
Notes: Current HDTV increases the field of view from 10° to 30°.
Asbinocular vision subtends over 120°, the next generation of
television aims toincrease the field of view to around 100°
(Austerberry 2011).
Figure 7STUDY 3 MANIPULATION AND STIMULI
A. Horizontal Display
B. Vertical Display
8 JOURNAL OF MARKETING RESEARCH, Ahead of Print
-
difference between horizontal and vertical conditions
inprocessing fluency.
We hypothesize that the initial processing fluency advan-tages
will increase perceptions of variety in the horizontalversus
vertical assortments. Because people generally donot seek
disconfirming evidence once perceptions have beenformed, we believe
that these increased perceptions of varietymay linger in the 15-sec
condition even after processing flu-ency advantages dissipate. It
is an empirical question whetherwe see the perceived variety effect
disappears as the process-ing fluency advantages fade away (i.e.,
an interaction effect)or we see a main effect of perceived variety
across bothtime conditions, suggesting that the initial perceptions
ofvariety are lingering. If the perceptions of variety linger
evenslightly, we should observe that participants choose a
largernumber of unique items in horizontal versus vertical
displays.
A 2 × 2 ANOVA conducted on the processing fluencyindex revealed
significant main effects of assortment ori-entation and exposure
time: participants found processingthe horizontal assortment to be
more fluent than processingthe vertical assortment (Mhorizontal =
5.86, SEM = .17 vs.Mvertical = 5.33, SEM = .17; F(1, 211) = 4.8, p
= .03, h2 =.02), and found the processing to be more fluent when
therewas more time (M3 sec = 5.32, SEM = .18 vs. M15 sec = 5.88,SEM
= .17; F(1, 211) = 5.3, p = .02, h2 = .02). A significanttwo-way
interaction (F(1, 211) = 4.0, p < .05, h2 = .02)was also found.
The planned comparisons showed that thedifference between
horizontal and vertical assortments inprocessing fluency was
significant in the 3-sec condition(Mhorizontal = 5.83, SEM = .25
vs. Mvertical = 4.80, SEM =.25; t(211) = 2.9, p = .004, h2 = .04)
but not in the 15-seccondition (Mhorizontal = 5.90, SEM = .24 vs.
Mvertical = 5.85,SEM = .23; t(211) = .1, p = .89, h2 = .0001). Put
differently,participants found processing the horizontal
assortmentto be equally fluent in the 3-sec and the 15-sec
conditions(t(211) = .2, p = .84, h2 = .0002) but found processing
thevertical assortment to be less fluent when time was
limited(t(211) = 3.1, p = .003, h2 = .04).
The same ANOVA conducted on the perceived varietyscale showed a
significant main effect of orientation(F(1, 211) = 6.0, p = .02, h2
= .03) but a nonsignificantorientation × time interaction (F(1,
211) = 2.7, p < .10, h2 =.01). Overall, participants reported
that there was morevariety in the horizontal (M = 7.14, SEM = .15)
than inthe vertical (M = 6.61, SEM = .15) assortment. Thus,initial
perceptions of variety appear to linger even whenmore time is
available. These lingering perceptions ofassortment variety should
then affect subsequent choice.
When we analyzed the number of unique options chosen,we found
only a significant main effect of assortment ori-entation, such
that participants indicated a higher number ofunique options for
the horizontal than for the vertical as-sortment (Mhorizontal =
3.98, SEM = .13 vs. Mvertical = 3.60,SEM = .13; F(1, 211) = 4.1, p
= .04, h2 = .02).
Our theory suggests that processing fluency is mediatingthe
perceived variety effect, which in turn mediates the choiceof
unique options. We ran the mediation analyses follow-ing the
recommendations by Zhao, Lynch, and Chen (2010).We used the SAS
macro provided by Preacher and Hayes(2004), and our bootstrap
analyses supported our proposedprocesses. First, the indirect
effect of assortment orientation(horizontal vs. vertical) on
perceived variety through the
hypothesized mediator, processing fluency, was significant(b =
.03, SE = .02; 95% CI = [.001, .07]). Second, the indirecteffect of
assortment orientation (horizontal vs. vertical) onnumber of unique
options through the hypothesized medi-ator, perceived variety, was
also significant (b = .04, SE = .02;95% CI = .005 to .08).
EXTENSION OF SHOPPING MALL FIELD STUDY
Study 3 suggests that participants chose more variety inthe
horizontal display because they perceived the horizontaldisplay to
have more variety due to horizontal processingadvantages. These
results shed light onwhy in our field studiespeople chose more
variety from horizontal versus verticalassortments. Although it
would be difficult to measure pro-cessing fluency advantages in a
shopping mall because thatwould require a tightly constrained time
frame for consumersto view the assortment, which would be
unrealistic, we couldmeasure whether or not there were lingering
perceptions ofincreased variety for horizontal versus vertical
displays.
For an extension of Study 3, we returned to the mall
en-vironment where we conducted the first field study. Wesurveyed
46 customers (45 females; age not collected), whoshopped from
either the horizontal or the vertical display,with the following
item: “Please estimate howmany differentfragrances of Pocket-Size
Hand Sanitizers are shown in thatstore display: _____ (please write
down a number between 10and 50).” Consistent with the lab study,
even after initialperceptual fluency advantages had presumably
dissipated,consumers perceived more variety in the horizontal (M
=32.14, SEM = 1.59) than in the vertical (M = 22.46, SEM =1.53)
display (F(1, 44) = 19.2, p < .0001, h2 = .30).
Study 3 demonstrates that horizontal (vs. vertical) as-sortments
increase processing fluency. That study and thefollow-up mall study
show that even after the processingfluency advantages of horizontal
displays dissipate withgreater processing time, perceived
assortment variety andsubsequently the number of unique options
chosen remainhigher for horizontal (vs. vertical) displays. In
Study 4, weuse eye-tracking apparatus to provide further evidence
thathorizontal (vs. vertical) assortments increase processing
effi-ciency when time is constrained. Moreover, when time
isunconstrained, the higher perceived variety of horizontaldisplays
should result in larger consideration sets, as wellas a more
extensive processing of displayed options.
STUDY 4: EYE-MOVEMENT EVIDENCE FORPROCESSING EFFICIENCY
ANDPROCESSING EXTENSIVENESS
Study 4 uses eye-tracking methodology to find attentionmeasures
to support our hypothesis that horizontal (vs. ver-tical)
processing is more efficient. These attention measures(e.g.,
fixations) allow us to develop an index to assess “pro-cessing
efficiency.” Specifically, we define processing effi-ciency as the
number of options fixated on divided by totalfixation time. It is a
ratio index and provides a behavioralmeasure to support the
perceived processing fluency scalesused in Study 3. We can also use
this methodology to confirmthat horizontal and vertical product
assortments indeed leadto primarily horizontal and vertical eye
movements, respec-tively. Finally, this study differs from the
previous studiesbecause we manipulate horizontal versus vertical
assortmentthrough physical layout of the assortment as well as a
graphic
Effects of Horizontal Versus Vertical Display 9
-
cue that induces horizontal or vertical eye movements.
Thisgraphic cue is consistent with a website design techniquecalled
“highlighting” (i.e., visually accentuating informationthrough
lines, colors, and shading; other techniques includesorting,
grouping, trimming, etc.; see Shi, Wedel, and Pieters2013). The
stimulus design is also consistent with commonlyseen retail
assortment sets (see Figure 8; see also Figures 1 and 2)
Method
Twenty students participated in the study for payment.Sample
size was determined according to previous researchand the
difficulty in time and effort in collecting this typeof data. A lab
assistant ran all participants, one at a time.She seated each
participant in front of a computer equippedwith a SensoMotoric
Instruments RED-m eye-tracking device,calibrated the participant so
the eye movements could betracked by the software, and continued to
monitor the partici-pant throughout the study. Each session lasted
about 30 min-utes, and each participant was paid $10.
The study consisted of two parts: a constrained-time con-dition
and an unconstrained-time condition executed withinsubject. If
horizontal displays are easier to process, then in
aconstrained-time condition, people should process more
al-ternatives per second of processing time. That is, they shouldbe
more efficient as measured by the number of alternativesfixated on
per second in the horizontal versus vertical condi-tions. If the
initial processing efficiencies yield lingering
perceptions of variety, then in the unconstrained-time
con-ditions, participants should spend more time and processmore
alternatives before making their decision. We canobserve
“processing extensiveness” by measuring totalprocessing time and
total number of alternatives fixatedon prior to making the final
choice.
Participants were told to quickly browse an assortmentof
lollipops available at www.melvillecandycompany.com.To make the
task more believable, we showed them a screenshot of the actual
website. Participants were also told theywould only be able to see
the assortment for a restrictedamount of time (10 sec).
Participants then viewed on thenext screen an assortment of 35
different lollipops, hori-zontally or vertically oriented (see
Figure 8, Panel A). Theireye movements during the 10-sec period
were recorded.We then removed the assortment from their vision and
dis-played the following response item: “Imagine you are in acandy
store and see this assortment of 35 lollipops. If youwant to buy
some for your own consumption, how manydifferent types of lollipops
would you like to purchase? Pleaseprovide a number between 1 and
35.”
Participants then proceeded to the second part of thestudy. From
the instruction, they learned that they would beshown the same
lollipop assortment again, and their taskthis time was to choose
which options they would want topurchase. They could pick as many
options as they wished,and they could look at the assortment as
long as they needed
Figure 8STUDY 4 (EYE-TRACKING STUDY) MANIPULATION AND
STIMULI
B: Retail Assortment Displays Similar to the (Vertical)
Assortment Stimuli
A: Horizontal Versus Vertical Assortment Stimuli
10 JOURNAL OF MARKETING RESEARCH, Ahead of Print
http://www.melvillecandycompany.com
-
to. Their eye movements were again recorded. Afterward,they
proceeded to the next screen and wrote down simpledescriptive terms
(“apple,” “butterfly,” “smiley face,” etc.)to indicate their
lollipop choices.
Results and Discussion
The lollipop assortment was delineated into specificareas of
interest (AOIs), with each AOI associated with asingle lollipop. We
analyzed the data such that all pointswithin an AOI were considered
a fixation on that specificlollipop.
Manipulation checks. To provide some evidence that par-ticipants
engaged in horizontal (vs. vertical) eye scanningwhen
processing a horizontal (vs. vertical) assortment, we firstneed
to categorize the seemingly random eye movements(see Figure 9,
Panel A) as either horizontal or vertical. Forany movement between
options, whereby the participant’seye fixations moved from one
lollipop AOI to another lol-lipop AOI, we assigned x1 and y1 to the
beginning pointof the movement and x2 and y2 to the end point of
themovement, with x and y representing the column and rownumber,
respectively, in the assortment (see Figure 9, PanelB). We then
calculated Dx (i.e., x2 − x1) and Dy (y2 − y1), aswell as the
difference between the magnitudes of Dx and Dy(i.e., |Dx| − |Dy|).
If an eye scan traveled a greater distance inthe horizontal than in
the vertical direction (i.e., |Dx| > |Dy|),
Figure 9DEFINING HORIZONTAL VERSUS VERTICAL EYE MOVEMENTS
A: Eye Scans of a Participant
B: Eye Scans Categorized as Horizontal or Vertical Eye
Movement
Δx = 4)
Δy = 1)
(x2 = 6, y2 = 3)
(x1 = 2, y1 = 2)
Horizontal
Δx = 1)
Δy = 4)
(x1 = 2, y1 = 2)
(x2 = 3, y2 = 6)
Vertical
Effects of Horizontal Versus Vertical Display 11
-
then we define it as a horizontal eye movement (left side
ofFigure 9, Panel B). If it traveled a smaller distance in
thehorizontal than in the vertical direction (i.e., |Dx| <
|Dy|),we categorize it as a vertical eye movement (right side
ofFigure 9, Panel B). If |Dx| = |Dy|, we consider the scan aneutral
eye movement.
Table 1 summarizes the frequencies of horizontal versusvertical
eye movements and reports SEMs and effect sizes(h2). In the first
part of the study (i.e., browsing stage/constrained-time
condition), we found significantly morehorizontal movements in the
horizontal condition (M = 16.38)than in the vertical condition (M =
5.27; F(1,17) = 39.8,p < .0001), significantly more vertical
movements in thevertical condition (M = 11.36) than in the
horizontal condition(M = 3.00; F(1,17) = 35.6, p < .0001), and
no difference in thenumber of neutral movements between conditions
(p = .15).These results show that horizontal (vertical) eye
movementsclearly dominated the processing of horizontal
(vertical)assortment.
In the second part of the study (i.e., choosing
stage/unconstrained-time condition), we observed similar pat-terns.
We found significantly more horizontal movementsin the horizontal
condition (M = 14.00) than in the verticalcondition (M = 4.36; F(1,
18) = 19.2, p = .0004), signif-icantly more vertical movements in
the vertical condition(M = 9.27) than in the horizontal condition
(M = 4.67;F(1, 18) = 14.8, p = .001), and again no difference
betweenconditions in the number of neutral movements (p =
.57).These results show that our manipulation of assortment
ori-entation successfully resulted in systematically horizontal
andvertical eye movements. These results are also consistent
withprevious findings that the layout of visual displays affects
thedirection of eye movements in both scene viewing (Tatler
andVincent 2008) and brand search task (Van der Lans, Pieters,and
Wedel 2008).
Note that a possible account for our hypothesis (i.e.,horizontal
scanning is more fluent and efficient than ver-tical scanning) is
that Western, English-speaking partici-pants are more familiar with
horizontal scanning becausethey are trained to read from left to
right (and not top tobottom, etc.). If this were the underlying
process, weshould see more left-to-right scanning than
right-to-leftand more horizontal scanning overall regardless of
assort-ment orientation. We conducted additional analysis on
theleftward/rightward or upward/downward direction of eyemovements
and found that there was no particular directionthat dominated
horizontal or vertical eye movements, thusruling out a “reading”
alternative hypothesis (for details, seeWeb Appendix A).
Processing efficiency in constrained-time condition. Table
2summarizes our main findings from hypothesis testing andreports
all SEMs and effect sizes (h2). In the constrained-time condition,
we predicted that the number of optionsfixated on per second, which
defines “processing efficiency,”should be greater in the horizontal
(vs. vertical) display con-dition. We found support for this
prediction. Participantsin the horizontal condition fixated on a
greater number ofoptions per second of processing time (M = 3.26)
thanthose in the vertical condition (M = 2.77; F(1,17) = 6.9, p
=.02). This processing efficiency advantage appears to bedriven by
both a directionally greater number of optionsfixated on in the
horizontal (M = 22.50, or 64% of totaloptions) than in the vertical
condition (M = 20.91, or 60%;F(1,17) = 1.0, p = .34) and a
marginally shorter total fixationtime (i.e., the total time spent
on fixated options; in seconds)in the horizontal (M = 6.89) than in
the vertical condition(M = 7.59; F(1,17) = 2.9, p = .10). That is,
participantsin the horizontal condition needed less time to fixate
onany given option, on average, than those in the
verticalcondition.
Table 1STUDY 4 (EYE-TRACKING STUDY) MANIPULATION CHECKS
Browsing/Constrained Time (n = 19) Choosing/Unconstrained Time
(n = 20)
Horizontal Display Vertical Display Horizontal Display Vertical
Display
Number of Eye Movements 21.25 (1.20) 19.36 (1.02) 20.67 (1.59)
16.18 (1.44)F-test F(1, 17) = 1.4 F(1, 18) = 4.4p-value .25 .05h2
.08 .20
Number of HorizontalMovements
16.38 (1.34) 5.27 (1.14) 14.00 (1.63) 4.36 (1.47)
F-test F(1, 17) = 39.8 F(1, 18) = 19.2p-value < .0001 .0004h2
.70 .52
Number of VerticalMovements
3.00 (1.07) 11.36 (.91) 4.67 (.89) 9.27 (.80)
F-test F(1, 17) = 35.6 F(1, 18) = 14.8p-value < .0001 .001h2
.68 .45
Number of NeutralMovements
1.88 (.43) 2.73 (.37) 2.00 (.70) 2.55 (.63)
F-test F(1, 17) = 2.3 F(1, 18) = .3p-value .15 .57h2 .12 .02
Notes: SEMs are reported in parentheses. We excluded one
participant’s data collected in the browsing stage/constrained-time
viewing condition becausea programming error led to a significantly
shorter stimulus exposure time for this participant.
12 JOURNAL OF MARKETING RESEARCH, Ahead of Print
-
After browsing the assortment for 10 sec, participantscreated
significantly larger consideration sets in the hor-izontal (M =
7.11 unique options in the consideration set)than in the vertical
condition (M = 3.18; F(1, 18) = 15.2,p = .001).
Processing extensiveness in unconstrained-time condition.In the
unconstrained-time condition, we predicted that largerconsideration
sets in the horizontal (vs. vertical) display con-dition should
lead to more extensive processing. We foundsupport for the expected
effects on processing extensiveness.First, the number of options
fixated on was greater in thehorizontal condition (M = 22.33, or
64% of all options) than inthe vertical condition (M = 17.73, or
51%; F(1, 18) = 4.7, p =.04). Second, the total fixation time (in
seconds) was longerin the horizontal condition (M = 12.20) than in
the verticalcondition (M = 7.85; F(1, 18) = 5.8, p = .03). As
expected,there was no difference between conditions on the
efficiencyof processing, that is, on the number of options fixated
on persecond (p = .27), suggesting that the processing
efficiencyadvantage afforded by the horizontal display dissipates
whentime is not constrained.
Participants engaged in more extensive processing of as-sortment
options and formed larger choice sets in the hori-zontal (M = 6.33
unique options in the choice set) versus thevertical condition (M =
3.91; F(1, 18) = 9.5, p = .006). Thisreplicates the results we saw
in Study 1, in which consumerswere allowed to purchase as many
items as they wanted(whereas in Studies 2 and 3, participants were
constrainedto a maximum number of options they could choose).
Our mediation analysis showed that, first, considerationset size
mediated the effect of horizontal versus vertical
display on processing extensiveness (b = 1.24, SE = .37,95%CI =
[.48, 2.00] for number of options fixated; b = 1.15,SE = .40, 95%
CI = [.37, 1.93] for total fixation time).Second, processing
extensiveness mediated the effect ofdisplay on choice set size (b =
.63, SE = .31, 95% CI =[.03, 1.24] for number of options fixated; b
= .72, SE = .33,95% CI = [.07, 1.37] for total fixation time).
These resultssuggest that horizontal (vs. vertical) displays, which
areeasier to process when time is constrained, can producelarger
consideration sets that in turn lead to more ex-tensive processing
and larger choice sets when time isunconstrained.
To summarize, this eye-tracking study shows that,first, the
horizontal and vertical assortment displays(created using both
physical layouts and graphic cues)indeed led to dominantly
horizontal and vertical eyemovements, respectively. Second,
scanning the horizon-tal (vs. vertical) assortment was more
efficient, character-ized by more options fixated on per second of
processingtime in constrained-time viewing. This processing
effi-ciency presumably resulted in higher perceived assort-ment
variety that produced larger consideration sets in thehorizontal
(vs. vertical) condition. However, in unconstrained-time viewing,
the horizontal processing efficiency evap-orated, but because
lingering perceptions of higher varietyand larger consideration
sets had been formed in thehorizontal condition from the initial
viewing, partici-pants spent more time and fixated on more options
whilechoosing from the horizontal (vs. vertical) assortment.
Thisresulted in larger choice sets in the horizontal (vs.
vertical)condition.
Table 2STUDY 4 (EYE-TRACKING STUDY) RESULTS
Browsing/Constrained Time(n = 19)
Choosing/Unconstrained Time(n = 20)
Horizontal Display Vertical Display Horizontal Display Vertical
Display
Processing Efficiency(Number of Options Fixated on per
Second)
3.26 (.14) 2.77 (.12) 2.02 (.18) 2.29 (.16)
F-test F(1, 17) = 6.9 F(1, 18) = 1.3p-value .02 .27h2 .29
.07
Processing Extensiveness(Total Number of Options Fixated On)
22.50 (1.23) 20.91(1.05) 22.33 (1.58) 17.73 (1.43)
F-test F(1, 17) = 1.0 F(1, 18) = 4.7p-value .34 .04h2 .05
.21
Processing Extensiveness (Total Fixation Time; sec) 6.89 (.31)
7.59 (.27) 12.20 (1.35) 7.85 (1.22)F-test F(1, 17) = 2.9 F(1, 18) =
5.8p-value .10 .03h2 .15 .24
Number of Unique Options in the Consideration Set 7.11 (.75)
3.18 (.68)F-test F(1, 18) = 15.2p-value .001h2 .46
Number of Unique Options in the Choice Set 6.33 (.58) 3.91
(.53)F-test F(1, 18) = 9.5p-value .006h2 .35
Notes: SEMs are reported in parentheses.
Effects of Horizontal Versus Vertical Display 13
-
STUDY 5: QUASI FIELD STUDY TESTINGBOUNDARY CONDITION
In the last study, we identify a boundary condition for
ourresults. In all of our previous studies, we maintained that
thereason more variety and more options were chosen (if thenumber
of options chosen was allowed to vary) in the hori-zontal
conditions than in the vertical conditions is that theprocessing
fluency associated initially with horizontal condi-tions yields
higher perceptions of variety. Those perceptionsof variety result
in larger consideration sets that eventuallylead to larger choice
sets (when quantity purchased is notlimited) or to more unique
items chosen (when choice setsizes are limited). However, in each
of these experiments,participants considered multiple-option choice
sets. If thechoice goal were to choose only one option, the focus
shouldbe on finding one’s favorite rather than identifying
accept-able variety. In that case, perceptions of variety should
notbe relevant and we should not see a difference between
hor-izontal and vertical displays, especially in situations in
whichpreferences are known.
In Study 5, we ask participants to make choices of well-known
Halloween candies in a laboratory. Pretests (N = 36)show that
participants have clear preferences among thebrands. The mean
preference scores (0 = “I don’t like it atall,” and 100 = “I like
it extremely”) for Hershey’s, KitKat,M&M’s, Snickers, and Twix
were 58, 66, 65, 31, and 34(SEMs= 4.61, 4.21, 4.85, 4.60, 5.73),
respectively (F(4, 140) =12.8, p < .0001, h2 = .23). Given that
Study 5 is a quasi fieldstudy, we asked participants to make real
purchases, but wedid so in a lab so that we could get additional
measures, suchas ease of choice task, satisfaction, and
confidence.
Method
Study 5 was a 2 (choice task: single vs. multiple options) ×2
(choice display: horizontal vs. vertical orientation)
between-subjects design. Our sample consisted of 271
undergraduatestudents who participated in the study in exchange for
coursecredit. Sample size was determined by recruiting as
manyparticipants as possible in one day in our lab, with a min-imum
requirement of 20 participants per condition. Par-ticipants were
told as a cover story that the study was a testof how store
lighting affects people’s shopping patterns. Tomake the task more
realistic, the lab was being designed asa ministore, and they would
be given $2 to spend in thestore. We set up four ministores, each
corresponding to anexperimental condition. In each store, each of
the five candieswas presented in bulk to the participants in five
transparentcontainers that could be arranged on a table either
verticallyor horizontally (similar to the Halloween field study
setup).We presented the candies in two random orders, which
werecounterbalanced across participants. Each candy was pricedat 25
cents per piece, and participants were told they couldpurchase as
many pieces as they had money for but had topurchase at least one
piece. They could keep any unusedmoney at the end of the study.
In the single-option condition, participants were told tochoose
their favorite type of candy and to decide how muchthey wanted of
that type. In the multiple-option condition,participants were told
to choose their favorite types ofcandy and to decide how many
pieces of each type theywanted to purchase. The candy containers
were refilled so
that every time a participant approached the containers,they all
appeared equally full.
We measured six dependent variables. First, we measuredhow much
time (in seconds) was spent making the choice.We also measured
purchase quantity (1–8) and number ofunique candies chosen (1–5).
Finally, to further understandthe underlying choice process, we
included three scale mea-sures adapted from the previous research
(Iyengar and Lepper2000). We hypothesized that when multiple
choices wereallowed, then participants would not have to make
trade-offs to determine their favorite, and they could choose
asmany varieties as theywanted. Thus, if horizontal (vs.
vertical)displays yield more perceived variety, we would expect
moretime spent on processing the assortment, as well as
greaterchoice ease (Townsend and Kahn 2014), choice
satisfaction,and certainty (Hoch, Bradlow, and Wansink 1999) in
thehorizontal (vs. vertical) condition for the
multiple-choicedecision. When participants needed to make a single
choice,the increased perceived variety would no longer be
relevant,and participants would focus on finding their favorite. In
thiscase, having to make the trade-offs to find a favorite wouldbe
more stressful (Luce 1998,) and, if anything, the higherperceived
variety in the horizontal (vs. vertical) conditionwould be a
liability. The three scale variables we includedto test these
hypotheses are as follows:
• Choice ease (1 = “not at all,” and 9 = “extremely”; a =
.62):1. Did you find it easy to make your decision of which
candy/candies to pick?
2. How much did you enjoy making the decision?3. How frustrated
did you feel when making the decision?(reverse-coded)
• Choice satisfaction (1 = “not at all,” and 9= “extremely”;a =
.89):1. How much do you like the candy/candies you have chosenfor
yourself?
2. How satisfied do you think you would be if you eat
thecandy/candies you chose?
3. How confident are you that the candy/candies will satisfy
you?• Choice certainty (1 = “not at all,” and 9 = “extremely”;a =
.56):1. How certain are you that you have chosen the best
candy/candies from the set?
2. How certain are you that the candy/candies will be among
thebest you’ve ever had?
3. How much do you regret choosing the candy/candies youpicked?
(reverse-coded)
Results and Discussion
We found no significant differences in purchase quantityamong
the four conditions. The overall mean number of itemspurchasedwas
4.03. In hindsight,we think that our instructionsmight have
suggested a normal purchase quantity of 4 becausewe provided the
following example in explaining how thepayout would work: “For
example, if you decide to buy4 pieces of candy, then you will spend
$1 (out of $2) andthen get the leftover $1.”
In spite of that, we do find our expected effects with thenumber
of unique items chosen. Obviously, in the single-choice condition,
there was no variance; by design, eachparticipant chose only one
type of candy. But in themultiple-choice condition, participants
chose significantly more uniqueoptions in horizontal (M = 3.00, SEM
= .10) versus verticalcondition (M = 2.36, SEM = .10; t(267) = 4.6,
p < .0001,h2 = .07).
14 JOURNAL OF MARKETING RESEARCH, Ahead of Print
-
A 2 × 2 ANOVA conducted on time (in seconds) spenton making the
choice revealed a significant main effect ofdisplay (F(1, 263) =
8.3, p = .004, h2 = .03), such that parti-cipants spent more time
choosing in the horizontal condition(M = 31, SEM = 1.44) than in
the vertical condition (M = 25,SEM = 1.45). This main effect is
consistent with our findingin the unconstrained-time condition of
the eye-tracking study.There was also a significant main effect of
task (F(1, 263) =6.2, p = .01, h2 = .02), such that people spent
more timechoosing in the multiple-option task (M = 30, SEM =
1.42)than in the single-option task (M = 25, SEM = 1.47).
Theinteraction between display and task was marginally signifi-cant
(F(1, 263) = 2.8, p = .09, h2 = .01), such that differencebetween
horizontal and vertical conditions was significant forthe
multiple-option choice (Mhorizontal = 35, SEM = 1.94 vs.Mvertical =
26, SEM = 2.08; t(263) = 3.3, p = .001, h2 = .04)but not for the
single-option choice (Mhorizontal = 26, SEM =2.13 vs. Mvertical =
24, SEM = 2.02; t(263) = .8, p = .40, h2 =.003). We excluded four
participants when analyzing this de-pendent variable because the
lab assistant did not record thetime these four participants spent
making the choice.
These results support the previous findings that
horizontaldisplays are perceived to have more variety and
requiremore time to choose among the options. The effect doesnot
disappear in the single-option condition (i.e., the interactionis
only marginally significant), but it is certainly mitigated
whenconsumers are focusing on a single-option choice. The
differ-ences between the two task conditions become clearer whenwe
study the scale measures.
The same ANOVA performed on the choice satisfactionindex showed
no significant main effect of task (p = .42) ordisplay (p = .33),
but we found a significant task × displayinteraction (F(1,267) =
6.7, p = .01, h2 = .02). Plannedcomparisons showed that for the
multiple-option task, thehorizontal display was significantly more
satisfying (M =7.92, SEM = .15) than the vertical display (M =
7.37, SEM =.16; t(267) = 2.6, p = .01, h2 = .02). For the
single-option task,the difference in satisfaction was not
significant between hori-zontal (M = 7.65, SEM = .16) and vertical
displays (M = 7.89,SEM = .15; t(267) = −1.1, p = .26, h2 =
.0005).
The findings were similar for choice confidence. Again,there
were no significant main effects, but the interactionwas
significant (F(1, 267) = 4.9, p = .03, h2 = .02). For
themultiple-option task, the horizontal display (M = 7.18,SEM =
.14) led to higher confidence in choice than thevertical display (M
= 6.60, SEM = .14; t(267) = 2.9, p =.004, h2 = .03). For the
single-option task, the differencewas not significant between
horizontal (M = 6.83, SEM =.15) and vertical displays (M = 6.88,
SEM = .14; t(267) = −.3,p = .79, h2 = .0003).
For the choice ease index, there was a significant maineffect of
task (F(1, 267) = 3.9, p < .05, h2 = .02), such thatthe
multiple-option task was seen as easier (M = 6.79,SEM = .12) than
the single-option task (M = 6.44, SEM =.13). Although the
interaction was only marginally sig-nificant (p = .10, h2 = .01),
within the multiple-option taskthe horizontal display (M = 7.04,
SEM = .17) was seen aseasier than the vertical display (M = 6.55,
SEM = .18;t(267) = 2.0, p = .04, h2 = .02), whereas within the
single-option task, there was no difference across horizontal (M
=6.40, SEM = .18) and vertical displays (M = 6.49, SEM =.17; t(267)
= −.3, p = .75, h2 = .0004).
Overall, these results show that when perceived varietywas a
relevant concern, as it is when choosing multipleoptions or more
than one flavor is the norm, horizontaldisplays yielded more unique
options chosen than verticaldisplays. In addition, in the
multiple-option task, partici-pants were more satisfied, were more
confident in theirchoices, and found a greater ease of making the
choicesthrough horizontal versus vertical displays. These
effectswere not evident in the single-option task, in which
par-ticipants focused on finding their favorite item. Whenpeople
choose a favorite option, variety is only relevantinsofar as it can
ensure that one’s favorite is present.
GENERAL DISCUSSION
Through a series of five studies, both lab and field
ex-periments, we have demonstrated that when choice situa-tions
allow for multiple options or multiple variants to beselected, more
variety is chosen (as measured by numberof unique options selected)
in horizontal displays than invertical displays. When the number of
options chosen is notfixed, it results in more options overall
being chosen fromhorizontal displays relative to vertical displays.
Through alab study and an eye-tracking study, we have found
supportfor at least part of the underlying processes that help
ex-plain why this effect occurs. During initial scans of
assort-ments, horizontal (vs. vertical) displays are easier to
processdue to a match between the binocular vision field (which
ishorizontal) and the dominant direction of eye movementsrequired
for processing horizontal displays. This process-ing fluency allows
people to more efficiently browse in-formation in a given time,
which increases the perceivedvariety of the horizontal (vs.
vertical) assortments. Althoughthese processing fluency advantages
dissipate with longerprocessing time, the perceptions of higher
variety linger,resulting in larger consideration sets in horizontal
(vs. vertical)assortments. More extensive processing is then needed
at thechoice stage, or when time is not constrained, to ruminate
onthe options, and this results in larger choice sets or moreunique
options chosen in horizontal (vs. vertical) assortments.We have
also identified a boundary condition whereby thedifferences between
horizontal and vertical displays evapo-rate during the choice of a
favorite single option amongfamiliar alternatives because perceived
variety is no longerrelevant or necessarily beneficial.
Because horizontal displays can make it easier for con-sumers to
scan the entire assortment and perceive the as-sortment variety, we
recommend horizontal assortmentdisplays be used in product
categories for which it is commonor desirable to purchase more than
one item within the cate-gory. In this case, horizontal displays
can increase varietyseeking and thus purchase quantities, as our
shopping mallfield study has shown. In product categories for which
con-sumers are more likely to make trade-offs and purchase onlyone
item, horizontal presentation of options might not be thebest
strategy. Although our current results show no differencein choice
ease, satisfaction, or confidence between horizontaland vertical
displays when only a single option is considered,we do see some
directional support that indicates verticalassortments might be
preferable in this context. For example,it is possible that
vertical assortments are more likely to cuehierarchical inferences,
making choosing one item easier. Weleave investigation of this idea
to future research.
Effects of Horizontal Versus Vertical Display 15
-
Our research contributes to several areas of research
inmarketing. First, the literature on consumer assortment
per-ceptions has shown that perceived assortment variety is
af-fected by various factors such as assortment structure
(Hoch,Bradlow, and Wansink 1999; Kahn and Wansink 2004), al-located
shelf space (Broniarczyk, Hoyer, andMcAlister 1998),presence of
high-preference items (Broniarczyk, Hoyer, andMcAlister 1998),
partitioning of the assortment (Mogilner,Rudnick, and Iyengar
2008), and visual versus verbal display(Townsend and Kahn 2014). We
add to this literature a newfactor: horizontal versus vertical
assortment display.
This literature has been interested in dependent variablessuch
as consumers’ store choice (Broniarczyk, Hoyer, andMcAlister 1998;
Hoch, Bradlow, and Wansink 1999), as-sortment preference (Townsend
and Kahn 2014), satisfaction(Hoch, Bradlow, and Wansink 1999;
Mogilner, Rudnick, andIyengar 2008; Morales et al. 2005), and
consumption quantity(Kahn and Wansink 2004). We add to this body of
work byshowing that horizontal versus vertical assortment display
canaffect processing fluency and efficiency, assortment
varietyperceptions, and size of consideration sets, which in turn
af-fects number of options and number of unique options chosen.
More broadly, our work contributes to the literature study-ing
the effects of in-store visual factors on consumer pro-cessing and
decision making. There, researchers have lookedat factors such as
shelf space (e.g., Inman andMcAlister 1993;for a review, see Campo
and Gijsbrechts 2005), number ofshelf facings (Chandon et al. 2009;
Drèze, Hoch, and Purk1994), and horizontal and vertical shelf
positions (Chandonet al. 2009; Drèze, Hoch, and Purk 1994;
Valenzuela andRaghubir 2009; Valenzuela, Raghubir, and Mitakakis
2013).Most related to our research is the work of Chandon et
al.(2009) and Valenzuela, Raghubir, and Mitakakis (2013)
onhorizontal and vertical shelf positions. Their findings showthat
a particular product’s position relative to other prod-ucts on the
shelf can influence how customers evaluate theproduct. Our work is
different in that it shows that whetheran array of products is
scanned horizontally or verticallyaffects processing fluency,
perceived variety, and choices.
Results from our studies provide important
managerialimplications. Although horizontal and vertical
assortmentdisplays as visual merchandising techniques are widely
andeasily used in retail settings, their effects on consumer
pro-cessing and decisionmaking are largely unknown tomanagers.Given
that assortment variety is one of the critical factors thatattract
consumers to a (physical or online) store (Broniarczykand Hoyer
2010), findings from this research suggest to re-tailers that
horizontal assortments can be used to promoteperceived assortment
variety.
In particular, our results suggest that horizontal
displaysshould be used when variety is a strategic aspect of
thecategory and when the retailer wants to make the varietymore
salient, which could then lead to increased purchasequantities. For
example, in candy or yogurt categories, forwhich variety is an
asset and multiple purchases are common,horizontal displays should
make the variety easier to processand should increase purchase
quantities. In these categories,different flavors are frequently
priced equivalently, which,again, should help encourage increased
purchased quanti-ties. In contrast, in purchase decisions in which
consumersare likely to choose only one item at a time, such as
whenprices or sizes vary, then vertical assortments should be
more
compelling because they de-emphasize variety. In these
cases,horizontal displays that allow consumers to process
morevariety might be confusing and thus make the choice amongthe
options more complicated.
As retailing increasingly moves to the mobile interface,there
may be further implications about what factors willmoderate our
results. In a mobile or tablet environment,product assortments can
easily be oriented either horizon-tally or vertically. Furthermore,
horizontal and vertical ori-entations can be simulated by viewing
products from eithera landscape or a portrait perspective as
consumers flip theirdevices from one angle to another. Future
research mightinvestigate whether our results would generalize to
these newtypes of retail platforms.
REFERENCES
Academic Medicine (1960), “Review of Medical Physiology
andBiophysics, edited by Theodore C. Ruch and John F.
Fulton,”Academic Medicine, 35 (11), 1067.
Austerberry, David (2011), “1080p and 3-D Developments,”
3DCineCast (blog), (December 9), [available at
http://3dcinecast.blogspot.com/2011/12/1080p-and-3-d-developments.html].
Broniarczyk, Susan M., and Wayne D. Hoyer (2010),
“RetailAssortment: More „ Better,” in Retailing in the 21st
Century,Manfred Krafft and Murali K. Mantrala, eds. Berlin:
Springer,271–84.
———, ———, and Leigh McAlister (1998), “Consumers’ Per-ceptions
of the Assortment Offered in a Grocery Category: TheImpact of Item
Reduction,” Journal of Marketing Research, 35(May), 166–76.
Campo, Katia and Els Gijsbrechts (2005), “Retail Assortment,
Shelf,and Stockout Management: Issues, Interplay, and Future
Chal-lenges,” Applied Stochastic Models in Business and Industry,21
(4/5), 383–92.
Chandon, Pierre, J. Wesley Hutchinson, Eric T. Bradlow, and
ScottT. Young (2009), “Does In-Store Marketing Work? Effects ofthe
Number and Position of Shelf Facings on Brand Attentionand
Evaluation at the Point of Purchase,” Journal of Marketing,73
(November), 1–17.
Cogan, David G. (1956), Neurology of the Ocular
Muscles.Springfield, IL: Charles C. Thomas.
Collewijn, H., Casper J. Erkelens, and R.M. Steinman
(1988),“Binocular Co-ordination of Human Vertical Saccadic Eye
Move-ments,” Journal of Physiology, 404 (1), 183–97.
Dickson, Peter R. and Alan G. Sawyer (1990), “The Price
Knowl-edge and Search of Supermarket Shoppers,” Journal of
Mar-keting, 54 (July), 42–53.
Drèze, Xavier, Stephen J. Hoch, and Mary E. Purk (1994),
“ShelfManagement and Space Elasticity,” Journal of Retailing, 70
(4),301–26.
DuPuis, Steven and John Silva (2008), Package Design
Workbook:The Art and Science of Successful Packaging. Beverly,
MA:Rockport.
Gilchrist, Iain D. and Monika Harvey (2006), “Evidence for a
Sys-tematic Component Within Scan Paths in Visual Search,”
VisualCognition, 14 (4/5/6/7/8), 704–15.
Griegel-Morris, Patricia, Keith Larson, KrissannMueller-Klaus,
andCarol A. Oatis (1992), “Incidence of Common Postural
Abnor-malities in the Cervical, Shoulder, and Thoracic Regions
andTheir Association with Pain in Two Age Groups of
HealthySubjects,” Physical Therapy, 72 (6), 425–31.
Hoch, Stephen, Eric T. Bradlow, and Brian Wansink (1999),
“TheVariety of an Assortment,” Marketing Science, 18 (4),
527–46.
16 JOURNAL OF MARKETING RESEARCH, Ahead of Print
http://3dcinecast.blogspot.com/2011/12/1080p-and-3-d-developments.htmlhttp://3dcinecast.blogspot.com/2011/12/1080p-and-3-d-developments.html
-
Hoyer, Wayne D. (1984), “An Examination of Consumer
DecisionMaking for a Common Repeat Purchase Product,” Journal
ofConsumer Research, 11 (3), 822–29.
Inman, J. Jeffrey and Leigh McAlister (1993), “A Retailer
Pro-motion Policy Model Considering Promotion Signal
Sensitivity,”Marketing Science, 12 (4), 339–56.
Iyengar, Sheena S. and Mark R. Lepper (2000), “When Choice
IsDemotivating: Can One Desire Too Much of a Good Thing?”Journal of
Personality and Social Psychology, 79 (December),995–1006.
Kahn, Barbara E. and Brian Wansink (2004), “The Influence
ofAssortment Structure on Perceived Variety and
ConsumptionQuantities,” Journal of Consumer Research, 30 (March),
519–33.
Kardes, Frank. R. (1986), “Effects of Initial Product Judgments
onSubsequent Memory-Based Judgments,” Journal of ConsumerResearch,
13 (June), 1–11.
Kröse, Ben J.A. and Bela Julesz (1989), “The Control and Speed
ofShifts of Attention,” Vision Research, 29 (11), 1607–19.
Lee, Angela Y. and Jennifer L. Aaker (2004), “Bringing the
Frameinto Focus: The Influence of Regulatory Fit on Processing
Flu-ency and Persuasion,” Journal of Personality and Social
Psy-chology, 86 (February), 205–18.
Luce, Mary Frances (1998), “Choosing to Avoid: Coping
withNegatively Emotion-Laden Consumer Decisions,” Journal of
Con-sumer Research, 24 (March), 409–33.
Mantrala, Murali K., Michael Levy, Barbara E. Kahn, Edward
J.Fox, Peter Gaidarev, Bill Dankworth, et al. (2009), “Why
IsAssortment Planning So Difficult for Retailers? A Frameworkand
Research Agenda,” Journal of Retailing, 85 (1), 71–83.
Metallinos, Nikos (1996), Television Aesthetics: Perceptual,
Cognitive,and Compositional Bases. New York: Routledge.
Mogilner, Cassie, Tamar Rudnick, and Sheena S. Iyengar
(2008),“The Mere Categorization Effect: How the Mere Presence
ofCategories Increases Consumers’ Perceptions of AssortmentVariety
and Outcome Satisfaction,” Journal of Consumer Re-search, 35
(August), 202–15.
Morales, Andrea, Barbara E. Kahn, Leigh McAlister, and SusanM.
Broniarczyk (2005), “Perception of Assortment Variety: TheEffects
of Congruency Between Consumers’ Internal and Re-tailers’ External
Organization,” Journal of Retailing, 81 (2),159–69.
Nazir, Tatjana A. (1992), “Effects of Lateral Masking and
SpatialPrecueing on Gap-Resolution in Central and Peripheral
Vision,”Vision Research, 32 (4), 771–77.
Ojanpää, Helena, Risto Näsänen, and Ilpo Kojo (2002),
“EyeMovements in the Visual Search of Word Lists,” Vision
Research,42 (12), 1499–512.
Preacher, Kristopher J. and Andrew F. Hayes (2004), “SPSS andSAS
Procedures for Estimating Indirect Effects in Simple Me-diation
Models,” Behavior Research Methods, Instruments, &Computers, 36
(4), 717–31.
Rayner, Keith (1975), “The Perceptual Span and Peripheral Cues
inReading,” Cognitive Psychology, 7 (1), 65–81.
——— (1998), “Eye Movement in Reading and InformationProcessing:
20 Years of Research,” Psychological Bulletin,124 (3), 372–422.
Reber, Rolf, Norbert Schwarz, and Piotr Winkielman
(2004),“Processing Fluency and Aesthetic Pleasure: Is Beauty in
thePerceiver’s Processing Experience?” Personality and
SocialPsychology Review, 8 (4), 364–82.
———, Pascal Wurtz, and Thomas D. Zimmermann (2004), “Ex-ploring
‘Fringe’ Consciousness: The Subjective Experience ofPerceptual
Fluency and Its Objective Bases,” Consciousnessand Cognition, 13
(1), 47–60.
Russo, J. Edward, Kurt A. Carlson, Margaret G. Meloy, and
KevynYong (2008), “The Goal of Consistency as a Cause of
Informa-tion Distortion,” Journal of Experimental Psychology:
General,137 (3), 456–70.
Shi, Savannah Wei, Michel Wedel, and F.G.M. Pieters
(2013),“Information Acquisition During Online Decision Making:
AModel-Based Exploration Using Eye-Tracking Data,”
ManagementScience, 59 (5), 1009–26.
Simmons, Joseph P., Leif D. Nelson, and Uri Simonsohn
(2011),“False-Positive Psychology: Undisclosed Flexibility in
DataCollection and Analysis Allows Presenting Anything as
Signif-icant,” Psychological Science, 22 (11), 1359–66.
Tatler, Benjamin W. and Benjamin T. Vincent (2008),
“SystematicTendencies in Scene Viewing,” Journal of Eye Movement
Research,2 (2), 1–18.
Townsend, Claudia and Barbara E. Kahn (2014), “The
‘VisualPreference Heuristic’: The Influence of Visual Versus
VerbalDepiction on Assortment Processing, Perceived Variety,
andChoice Overload,” Journal of Consumer Research, 40
(5),993–1015.
Tversky, Amos and Daniel Kahneman (1974), “Judgment Un-der
Uncertainty: Heuristics and Biases,” Science, 185
(4157),1124–31.
Valenzuela, Ana and Priya Raghubir (2009), “Position-Based
Beliefs:The Center-Stage Effect,” Journal of Consumer Psychology,19
(2), 185–96.
———, ———, and Chrissy Mitakakis (2013), “Shelf SpaceSchemas:
Myth or Reality?” Journal of Business Research,66 (7), 881–88.
Van der Lans, Ralf, Rik Pieters, and Michel Wedel (2008),
“EyeMovement Analysis of Search Effectiveness,” Journal of
theAmerican Statistical Association, 103 (482), 452–61.
Yeshurun, Yaffa and Marisa Carrasco (1999), “Spatial
AttentionImproves Performance in Spatial Resolution Tasks,”
VisionResearch, 39 (2), 293–306.
Zhao, Xinshu, John G. Lynch Jr., and Qimei Chen (2010),
“Mythsand Truths About Mediation Analysis,” Journal of
ConsumerResearch, 37 (August), 197–206.
Effects of Horizontal Versus Vertical Display 17
-
There are no queries in this article.
AUTHOR QUERIES
AUTHOR PLEASE ANSWER ALL QUERIES
A “Wide” Variety: Effects of Horizontal Versus Vertical Display
on Assortment Processing, Perceived Variety, and ChoiceEffe
...STUDY 1: SHOPPING MALL FIELD STUDYSTUDY 2: HALLOWEEN FIELD
STUDYPROCESSING OF INFORMATION IN HORIZONTAL VERSUS VERTICAL
DISPLAYSSTUDY 3: LAB STUDY MEASURING PROCESSING FLUENCY AND
PERCEIVED VARIETYMethodResults and Discussion
EXTENSION OF SHOPPING MALL FIELD STUDYSTUDY 4: EYE-MOVEMENT
EVIDENCE FOR PROCESSING EFFICIENCY AND PROCESSING
EXTENSIVENESSMethodResults and DiscussionManipulation
checks.Processing efficiency in constrained-time
condition.Processing extensiveness in unconstrained-time
condition.
STUDY 5: QUASI FIELD STUDY TESTING BOUNDARY
CONDITIONMethodResults and Discussion
GENERAL DISCUSSIONREFERENCES